Abstract Practical reasoning is often described as weighing reasons. When one deliberates about what to do one puts all the reasons for the action on one side and all the reasons against the action on the other side. The balance between both sides determines the outcome of the deliberation. Assuming that this description is correct, the next question is how the different reasons for and against the action determine the outcome of the deliberation. This is the place where the notion of weight enters. The natural answer is that every reason for or against an action has a weight, and the weights of all the reasons involved determine what one should do. The aim of this paper is to argue that this answer is wrong; weights of reasons have no role in a theory of reasoning. I. INTRODUCTION You are offered a job and you have various reasons to accept it, but before you give your answer you get another offer. The second job has the same advantages and disadvantages as the first one except that the salary is better. You decide to take the second job and the explanation for your easy decision can be put in terms of weight of reasons. In both options the salary was a reason to take the job, but in the second case this reason was weightier. Now, suppose that before accepting the second offer you get a third one. This time the salary is lower but the job is much closer to home, now the decision is less easy and you think it over. In the end you decide to take the third job. It is natural to explain your new decision in terms of weight of reasons again; the weight of the distance-reason was greater than the weight of the salary-reason. The aim of this paper is to show that the new explanation is an over generalization. This generalization lacks theoretical support; comparisons of weights of reasons (weighing reasons) cannot explain our less easy decisions. Practical reasoning is often described as weighing reasons. For example, in deliberating whether to take the third job mentioned above, you put on one side the reasons for one job (close to home) and on the other side you put the reasons against the job (the salary is not so high). The balance between both sides determined your decision. Assuming that this description is correct in general, the next question is how the different reasons determine the outcome of the deliberation. This is the place where the notion of weight enters. The answer in caricature is that every reason for or against an action has a numerical weight, so that the sum of the weights of the reasons for the action and the sum of the weights of the reasons against the action, determines what one should do. Most philosophers find this caricature absurd; still, many will say that if we do not take the idea of weight too literally, the description is adequate and it can explain how reasons determine what one should do. For example, Broome (2004) writes: Each reason is associated with a metaphorical weight. This weight need not be anything so precise as a number; it may be an entity of some vaguer sort. The reasons for you to φ and those for you not to φ are aggregated or weighed together in some way. The aggregate is some function of the weights of the individual reasons. The function may not be simply additive… It may be a complicated function, and the specific nature of the reasons may influence it. Finally, the aggregate comes out in favor of your φing, and that is why you ought to φ. Broome mentions two features of literal weight that are not necessarily true of metaphorical weights. First, a metaphorical weight is not necessarily a number; secondly, the weight of the whole is not necessarily the sum of the weights of the parts. So what is left from the idea of weight? Two minimal requirements are essential. First, an object is not identical to its weight; otherwise, the notion of weight of a reason does not add anything to the theoretical work of the notion of reason. Weights are features of objects that can be distinguished from them so that there might be distinct reasons that weigh the same. Secondly, the weight of a combination of objects is a function of the weights of the objects. In Broome's words: ‘The aggregate is some function of the weights of the individual reasons.’ In sum, if the metaphor of weight has a place in a theory of reasoning, the weight of a reason, is a feature of a reason which is distinct from it, and the following thesis is true: The evaluative judgment—what one ought to do in a specific situation—is determined by the weights of the reasons for and against that judgment. I believe that this is the substantial thesis that is left after the metaphorical features are ignored and I will call it the determination-by-weight thesis (DWT). The notion of weight of reasons is a theoretical one, it is supposed to explain how various and sometimes conflicting reasons determine the normative result. If DWT is false, the notion of weight cannot fulfil its theoretical role and this is tantamount to saying that there are no such entities as weights of reasons.1 This is what I aim to show in this paper. If the argument of this paper is sound, the notion of weight has no role in an account of reasoning, and weighing reasons has nothing to do with the weight of reasons.2 After explaining DWT (Section 2), I will show that if weights of reasons are invariant then DWT is false (Section 3). The rest of the paper will deal with the more complicated case where holism about reasons holds and the weight of a reason varies with context. In Section 4, I will distinguish between empty determination and substantive determination. In Section 5, I will show that the tension between holism and substantive DWT can be resolved only if a moderate account of our capacity to recognize the weight of reasons is available. In Section 6, I will present a moderate account of our capacity to recognize reasons. In Section 7, I will argue that we cannot offer such an account for our capacity to recognize the weights of reasons, concluding that if holism about reasons holds, DWT is empty. This establishes my case against DWT, since if holism is false so is DWT and if holism is true DWT is empty. In Section 8, I conclude by suggesting that in a realistic account of practical reasoning the notion of weight is not needed. II. EXPLANATION OF THE DWT The essential element in the notion of weight of reasons is its capacity to determine (with the weights of the other components) whether a certain action should be done. Broome is right in pointing out that determination is a function from features of the components to features of the whole and it does not have to involve numbers. The best example of a determination thesis that has nothing to do with numbers is the composition principle in semantics. The semantic value of a sentence is determined by the semantic values of its components and the way they are arranged (I will omit the last qualification in what follows). The composition principle applies both for the sense of expressions and for their references. For simplicity, I will focus on the version of compositionality which takes the semantic value of a sentence as its truth-value and the semantic value of the components as their references. The truth value of a sentence is determined by the references of its parts. In the same way, the normative value of the whole is determined by the normative values of its parts. In the practical case, we can take the normative value of the whole as the answer to the question whether to do A. According to DWT, this answer is determined by the normative values of the relevant features of the situation.3 The relevant features are reasons for and against doing A and the normative values of these features are the weights of the reasons. Semantic values, like normative values, are theoretical concepts; we do not know, before we develop the theory, what the references of words are, but we do know that whatever they are, if two words have the same reference there is no difference in the contribution they make to the truth value of the sentences in which they appear. The same is true for weights of reasons, we do not know, before we develop the theory, what the weights of reasons are, but we do know that if two reasons have the same weight there will be no difference in the contribution they make to the normative value of situations in which they appear. This analogy goes further, compositionality in semantics is supposed to explain how we understand a new sentence; similarly, DWT is supposed to explain how we decide what to do in a new situation. Given their explanatory role, compositionality and DWT are not only metaphysical. If references are entities which a competent speaker cannot recognize then even if the compositionality principle is correct, it is useless. Similarly, if weights of reasons are entities that a competent agent cannot recognize, then the fact that those weights determine the normative result is useless in that it cannot explain how an agent makes normative sense of a new situation. Hence, in criticizing DWT I do not understand it as an abstract metaphysical thesis, but as a thesis that should have implications for our understanding of practical reasoning. The semantic model was introduced to support Broome's claim that the idea of a ‘metaphorical weight’ can be substantiated without using numbers. But as is clear from the previous paragraph the analogy with the semantic model is much deeper; in what follows this analogy will have a crucial role in developing my argument against DWT. A few more clarifications are needed before I start my argument. First, the issue of determination by weight should be distinguished from the issue of incommensurability of values. The latter issue is about the order relation between possible outcomes, while the issue of this paper is how the order relation between two outcomes is determined by their components. The focus of the issue of incommensurability of values is comparison, while the focus of my concern is composition. In the example of the opening paragraph I do not deny that the three offers can be compared and even ordered, I can even accept that we can assign values4 (or utilities) to the three offers in a way that respects the order relations between them. Assuming that your decision to take the third job was correct, the value of the third job is higher than the value of the second job. Still the question of this paper remains unanswered: How the utility or value of the outcome of an action is determined by the reasons for and against the action? Secondly, assuming as I did that each outcome has a value, there is a metaphysical question about the grounds of those values. Are those values constituted by one's preferences or are one's preferences simply tracking objective values? This metaphysical question is outside the scope of this paper. The question of this paper is how (and if) the normative value of an outcome is determined by the normative values of its parts; and not the metaphysical grounds of those normative values. An answer to the metaphysical question might have implications for the issue of this paper and vice versa.5 But the implications are not straightforward; hence, it is important to separate the issues. Thirdly, since the weight of a reason is not identical to the reason itself, the DWT should be distinguished from what I will call determination-by-reasons thesis. The determination-by-reasons thesis is the claim that the right thing to do is determined by the reasons for and against the relevant action. I have no objection to this claim; my target is the idea that we can explain how reasons determine the normative result by the DWT. To understand the distinction between those two theses, compare the following two claims: The tastiness of a dish is determined by the tastes of the components. The tastiness of the dish is determined by the tastiness of the components. The first claim is approximately true,6 the second seems absurd. Even if avocado is as tasty as chocolate, salmon with avocado is tastier than salmon and chocolate. This is because the taste of avocado is different from the taste of chocolate even if one is as tasty as the other. The elements that determine the tastiness of the whole are the tastes of the ingredients, not their tastiness.7 I will argue that the same is true about reasons and their weights. What one ought to do is determined by one's reasons; but the DWT is similar to claim 2 in looking only at the tastiness (weights) of the ingredients and ignoring their other features. My argument against DWT is presented in the form of a dilemma. The first horn of the dilemma assumes that the weight of a reason is invariant (Section 3). The second horn accepts holism about reasons and their weights: a feature can be a (strong) reason for an action in one context without being a (strong) reason for this action in another context. I will argue that in both cases the DWT fails, in the first case it is false, and in the second case it is empty. III. IF HOLISM ABOUT REASONS IS FALSE DETERMINATION-BY-WEIGHT IS FALSE If holism about reasons is false, reasons are context-independent; Kagan (1988:12) calls this assumption the ubiquity thesis: if a feature makes a difference anywhere it must make a difference everywhere.8 The first horn of the dilemma is a stronger assumption, namely, that if A is a normative feature of a situation, it contributes the same normative weight to any other situation in which it appears. This assumption is stronger since it is possible that a reason is invariant, but its weight changes with context. For example, avoiding pain might be an invariant reason against an action, but in the context of physical exercise it is less weighty than in other contexts. In this section, I assume that weights of reasons are invariant, and in the next section I assume that the weight of a reason is context-dependent (whether the reasons themselves are invariant or not). One way to show that a determination thesis is false is by a counter-example of the following structure. A and B are two distinct reasons that have the same weight while the combination of A and C is weightier than the combination of B and C.9 I will suggest an example with the same structure to refute the DWT. The fact that Alex will be in the party is a good reason for me to join the party. The fact that Bruno will be in the party is also a good reason to join the party. Assume that those reasons weigh the same in this specific context; if I have to choose between a party with Alex and a party with Bruno, and there are no other considerations involved, I am indifferent. If the weight of the reason that Alex (Bruno) will be in the party is w(A)(w(B)); then w(A) = w(B). In another context, Carl comes to the party too, and as we all know, this might change the situation so that I am not indifferent anymore; I prefer a party with Bruno and Carl to a party with Alex and Carl.10 (This is similar to the avocado and the chocolate case.) Since the weight of a reason has to stay fixed, w(A) = w(B) in the new context as well. So, we have a counter example to the DWT: w(A+C) is different from w(B+C) although w(A) = w(B). The natural response to this example is that spending the evening with Alex (or Bruno) and Carl should not be decomposed to spending the evening with Alex (or Bruno) and spending the evening with Carl. It is better, for example to decompose spending the evening with Alex (or Bruno) and Carl to the contribution of Alex (or Bruno) to a party and the contribution of Carl to a party; Alex contributes jokes, Bruno contributes gossip and Carl contributes original ideas. Each of these contributions has a weight and all those weights will determine the weight of the reason to join a party with Alex and Carl. My answer is that the new decomposition faces the same problems as the original one, Alex's contribution to an evening might be similar in weight to that of Bruno, still my reason to spend the evening with Bruno and Carl might be weightier than my reason to spend the evening with Alex and Carl. Let's look more carefully at the natural response and my answer to it. We are comparing a party with Bruno and Carl to a party with Alex and Carl and the decomposition suggested is the following. An evening with A and C is composed of jokes and interesting ideas, an evening with B and C is composed of gossip and interesting ideas. Suppose that I like A’s jokes as much as I like B’s gossip, still I prefer the combination of gossip and interesting ideas to the combination of jokes and interesting ideas. The decomposition suggested still refutes DWT since w(jokes) = w(gossip) but w(jokes+interesting ideas) is different from w(gossip+interesting ideas). The objector might say that if I prefer B and C to A and C, there must be a further element that accounts for this preference, maybe C does not like jokes, and this creates tension every time that A tells a joke. If this is true, the objector continues, the tension should also be added to the balance of reasons; we have to add to an evening with A and C the (negative) weight of the tension created by the fact that C does not like A’s jokes. Hence, an evening with A and C is composed of jokes, interesting ideas and tension, while an evening with B and C is composed of gossip and interesting ideas. Our assumptions about the weights of the ingredients are enough to determine that the weight of an evening with B and C is greater than the weight of an evening with A and C. Hence, the outcome of the deliberation is determined by the weights of the reasons and my example does not refute the DWT. This objection rests on a wrong assumption; it is assumed that whenever an evening with B and C is better than an evening with A and C, it is because a new element was added to one of the options, such that the weight of this new element accounts for the fact that one of the options is weightier than the other. This might be true in the case that C does not like A’s sense of humour, and thus a new element of tension is added. But it is wrong to assume that it is always the case. Even if C likes jokes and no tension is added, I might prefer (whether this preference is tracking independent value or not) the combination of gossip and interesting ideas to the combination of jokes and interesting ideas. Let's return to the culinary example, avocado with salmon is tastier than chocolate with salmon not because a new taste was added to the combination of the salmon with chocolate. Compare this to another culinary example, spinach might be as tasty as potato, but spinach with wine is less tasty than potato with wine. Here it is true that another (metallic) taste was added to the combination of spinach and wine and this new non-tasty element explains why this combination is less tasty. But this is not what accounts for the failure of the determination of tastiness in the chocolate case. Some combinations of tastes are nicer than others as some combinations of features of a social evening are nicer than others. Of course, the fact that the determination thesis failed for specific decompositions does not prove that it will fail for all. The aim of my example was not to prove the impossibility of such decomposition, but to show how demanding the idea of determination by context-independent weights is. I cannot see anything that will justify the idea that there must exist a decomposition of the evening with A and C into a set of parameters such that only their weights as reasons will determine the force of the reason to spend the evening with A and C. It seems that the burden of proof is on the defender of the determination thesis. The objector might try to shoulder this burden by suggesting the following decomposition. The value (or normative weight) of an evening with A and C is determined by the value of an evening with C and the value of the contribution of A to an evening with C. Since the contribution of A to an evening with C is different from the contribution of B to an evening with C, the determination thesis is left intact. This might be true, but with this suggestion the objector is moving towards holism about reasons and their weights. So, let's move with him to the second horn of the dilemma which assumes that the weights of reasons are context-dependent.11 IV. EMPTY DETERMINATION VS SUBSTANTIVE DETERMINATION If the weights of reasons vary from one context of deliberation to another, we can trivialize the determination thesis by assigning weights to the different normative features according to the result of the deliberation. The holistic defender of DWT can reject any counter-example by playing with the weights of the reasons. He can reject the counter-example of the previous section by stipulating that w(A) = w(B) = 5 in the first context and in the second context where Carl is also present in the party w(A) changes to 2. Assuming that w(B) remained fixed and that w(C) = 7, we get that w(A+C) = 9 and w(B+C) = 12; this is indeed the desired result given that the final evaluative judgement is that in the new context I prefer a party with Bruno to a party with Alex. This is what I call empty determination. The use of numbers was an extreme simplification; my point can be made without assuming that weights of reasons are numbers. The above manoeuvre can also be applied to the following determination thesis which is obviously false. For example; the thesis that the meaning of a word is determined by the meanings of the letters is obviously false. But we can make it trivially true by assigning to the letter c in the word circle the meaning line; and to the letter i the meaning closed and so on. In the word cat the meaning of c is completely different, but this is not a problem if we accept that the meaning of a letter varies with context (holism about the meaning of letters). As long as there are no constraints on weights, holism implies an empty determination thesis. In order to have a substantive determination thesis that will be able to explain how the whole is determined by its parts; we must have constraints on the weights of reasons. These constraints must give priority to the parts in that their weight cannot depend on the value of the whole. On the other hand, holism seems to give priority to the whole. This is what creates the tension between substantive determination and holism. In the next section, I will present a necessary condition for the resolution of this tension. V. MODERATE ACCOUNT VS RADICAL ACCOUNT As already mentioned, the tension between holism and determination is an important issue in philosophy of language. Determination in semantics is expressed by the principle of compositionality, and holism in semantics is the thesis that the semantic value of a word changes with context.12 We can bring some of the considerations from the semantic case to bear on the tension in general.13 I will start by introducing a distinction between two ways to account for a speaker's capacity to understand a word in a new context. Both accounts are contextualist in the sense that they tell us how the context reveals the semantic contribution of a term in this context. They differ in what they demand the speaker to know about the context in order to know the semantic contribution of a term. If the speaker has to know the semantic value of the whole in order to know the contribution of the part, I will call the account radical account. If what the speaker has to know are features of the context that do not depend on the semantic value of the whole I will call the account moderate.14 The following examples will clarify this distinction. In the sentence ‘Bill is smart’ the reference of Bill depends on the context; however, the context determines which ‘Bill’ we are talking about independently of the truth value of the whole sentence. We don’t decide that we are talking about Bill A and not about Bill B because Bill A is smart and Bill B is not. The reference of ‘Bill’ depends on some features of the context: those features might be the ‘Bill’ we talked about a minute ago, or the ‘Bill’ that just left, or the ‘Bill’ that we all know well. It is probably impossible to give a closed list of those features; but it seems that a speaker can recognize those features without knowing the truth value of the whole sentence. Those features determine the reference of ‘Bill’ independently of the reference of the whole sentence. Hence, the account of the speaker's capacity to know the semantic value of ‘Bill’ in a specific sentence does not involve knowledge of the truth value of the sentence. The account is moderate. On the other hand, in the sentence ‘Bill dressed the chicken’ the reference of the word ‘dressed’ is not determined independently of the truth value of the sentence; we choose the interpretation (reference) under which the sentence is more likely to be true. Such a case calls for a radical account. On a moderate account (as in ‘Bill’), the value of the parts can be determined by the context before knowing the value of the whole. Hence, the value of the whole can be determined by those values without circularity and substantive determination still holds. However, if the only available account is radical (as in ‘dressed’), we have an empty determination. My point is reinforced by the observation that if the reference of every word depended on context, as in ‘dressed’, the compositionality thesis would lose its explanatory power completely.15 The lesson from the semantic case is that only a moderate account of one's capacity to recognize the contribution of each feature in a new context will resolve the tension between substantive determination and holism. Without a moderate account determination is empty in that it cannot explain how one's recognition of the value of the whole is determined by one's recognition of the values of the parts. In what follows it will be shown that although a moderate account of one's capacity to recognize reasons is available (Section 6), we cannot have such an account for one's capacity to recognize weights of reasons (Section 7). VI. A MODERATE ACCOUNT OF RECOGNIZING REASONS The moderate account that follows draws on the distinction between reasons and enablers. Both reasons and enablers are features of the situation that are relevant in determining what to do in that situation. The distinction between them is based on the idea that in order for a fact to function as a reason, some background conditions need to obtain. Those conditions are not reasons, since by themselves they do not count in favour of acting in a certain way; they only enable the relevant reason to count. These conditions are called enablers. Disablers are interfering conditions that when they obtain, the fact might not function as a reason. This distinction is suggested by Dancy and it plays a crucial role in his defence of Holism about reasons. According to Dancy, although the fact that I promised you X is generally a reason to do it, there are contexts (for example, where I was forced to promise) in which this same fact is not a reason to do X. In the new context, where I was forced to promise, the fact that I promised is not a reason at all. It is not that I have a reason to do X (I promised) and I have a conflicting reason not to do X (I was forced to promise). Defending this way to conceptualize such examples leads Dancy to the claim that reasons are context-dependent. Dancy's distinction can be used in developing a moderate account of our capacity to recognize context-dependent reasons.16 A competent reasoner is able to recognize a prima facie reason for an action in a specific situation. This ability is not enough; because what might be a reason in one context might not be a reason in another. The reasoner has to recognize enablers and disablers of this specific reason in this specific context. When the reasoner exercises this complex capacity, he sees how the normative contribution of each reason is influenced by the relevant enablers and disablers. This account meets ‘the challenge to explain how it is that an element retains its independent purport case by case, even though it makes different contributions to different contexts’ (Dancy 2004:198). The independent purport that an element has is explained by the fact that the (dis)enablers determine this normative contribution independently of the other elements and of the whole. This is only a very rough outline of an account. The most elaborate account along these lines is given by Horty who uses the framework of default logic in accounting for practical reasoning.17 The main tool in default logic is that of a default rule, a default rule is a defeasible rule that tells us what to believe or do as long as the rule is not defeated. A reason can be seen as an antecedent of such a rule. Without entering into the technical details, the idea is that when a competent reasoner recognizes a reason for an action, he is also aware of its possible defeaters (the defeaters are similar to Dancy's disablers). If one of its possible defeaters is present, the reason is defeated and is not relevant anymore.18 Ideally, at the end of this process all the reasons that favour one option are defeated so that all the undefeated reasons favour the other.19 The lesson which emerges from Horty's account is that a moderate account of our ability to recognize context-dependent reasons is available. This saves the determination-by-reasons thesis from the threat of Holism. However, as stressed from the start, my target is not the determination-by-reasons thesis, but the DWT. In the next section, it will be shown that Horty's moderate account of our capacity to recognize reasons cannot be extended to account for our supposed capacity to recognize the weights of reasons. VII. A MODERATE ACCOUNT OF WEIGHTS OF REASONS IS IMPOSSIBLE A reason is a consideration for or against an action. I will use the term direction of a reason to refer to the fact that it is for or against the action. So Holism can be reformulated as: the direction of a reason is dependent on context. 20 According to the thesis that I want to reject, reasons have not only directions, but also weights. If reasons do have weights, then according to Holism these weights also vary with context. Enablers and disablers explain the variability in the direction of a reason; while intensifiers and attenuators explain the variability in their weights (see Dancy 2004: 41–2). According to Horty's moderate account about reasons, we do not need to know the final evaluation in order to know the direction of a reason, we only need to recognize the enablers and disablers of the specific reason in the specific context. Analogously, it seems that we need to recognize only the intensifiers and attenuators of a reason in order to know its weight in any specific context. If a moderate account about weights can be developed from this analogy then the claim that reasons have context-dependent weights that determine what one should do in the specific context is true and non-empty. I will show that although such a moderate account about weights is similar in structure to Horty's moderate account about reasons, it cannot work. The important difference between these two accounts is due to a deep asymmetry between the notion of direction of a reason and the notion of weight of a reason. Horty's moderate account about reasons describes our reasoning as starting with prima facie reasons and then checking defeaters and defeaters of defeaters and so on. Since this model is recursive, it is crucial that there is a starting point; otherwise, we are in a vicious circle. The starting point is our recognition of prima facie or default reason. A rough definition of this notion is what I will call the isolation test.21 A prima facie reason for A is a feature that if there are no other considerations one should do A. A prima facie reason against A is defined similarly. As we saw in the previous section, the notion of prima facie reason is vital for Horty's account. It is important to notice further that this notion cannot be understood without the notion of the direction of a reason. Using the isolation-test, we start by defining a prima facie reason with a direction (for or against the action) and only after that we can define the notion of a prima facie reason: a prima facie reason is a prima facie reason for an action or a prima facie reason against an action. Without an understanding the notion of prima facie direction, we cannot understand the notion of a prima facie reason. This is not surprising as it is essential to the concept of a reason that it has a direction; a reason is a consideration for or against something. By analogy to the centrality of prima facie reasons in Horty's account; prima facie weights will be essential in a moderate account of weight. However, the notion of prima facie weight of a reason is highly problematic. The isolation-test cannot help us to understand it, since a reason which is isolated from all other reasons has only a direction towards a certain action; there is nothing to add that might correspond to weight.22 My criticism of the notion of prima facie weight is not just that we lack a definition for it, but that we lack understanding of it. To understand the notion of a prima facie weight of a reason we have to think of it in abstraction from other reasons. However, given the two following claims about weights this abstraction is useless. The first claim is that the notion of a weight of a reason is a comparative notion; it is based on comparisons between reasons (A is a stronger or weightier reason than B). The second claim is Holism about weights. If the weight of an object depends on the context of weighing, and we weigh objects by comparing them to other objects the question: ‘what is the weight of A when there are no other objects to compare it to?’ is meaningless. The only way to answer this question is by looking at the counterfactual situation where we put A next to another object and compare between them. This gives us the weight of A in the counterfactual situation. However, if Holism is true, the weight of A in the counterfactual situation tells us nothing about the weight of A in the original one. Without an understanding of the notion of prima facie weight we cannot build a moderate account of weight along the lines of the moderate account of reasons suggested by Horty. 23 One might object by claiming that we do not need the notion of prima facie reason (let alone the isolation test) to account for our capacity to recognize context dependent reasons. The advocate of DWT who accepts holism may simply say that we are capable of knowing what reasons there are in a case just by considering the case at hand. The notion of a reason is an ordinary notion; we can recognize the existence of reasons prior to accepting any theories or tests. In many cases, we can perfectly well know what the relevant reasons, enablers, disablers etc. are before knowing the final evaluation. I have some sympathy for the claim that since the notion of reason is ordinary and non-theoretical, the notion of prima facie reason (and Horty's account that is built on it) is not needed. But this only reinforces my point about the asymmetry between the notion of direction of a reason and the notion of weight of a reason. Although we do recognize some reasons as weighty and others as trivial, we do not recognize their weights. We do not have an ordinary notion of weight of a reason; we do not even know whether it is a number or some other kind of entity. In contrast to the notion of reason (and its direction) that might be clear enough without a theory, the notion of weight demands a theory. If the argument of this section is correct, the prospects for such a theory are dim and hence the notion of weight cannot fulfil its theoretical role. In the concluding section, I will suggest that my conclusion that reasons have no weight is less worrying than it might seem since the notion of weight is dispensable. VIII. CONCLUSION: WEIGHING WITHOUT WEIGHTS I have no objection to the claim that the right thing to do is determined by the reasons for and against the relevant action.24 My target was the claim that the right thing to do is determined by the weights of those reasons (DWT). DWT was supposed to explain how reasons determine the normative result. Hence, if my argument against DWT works, it seems that we are left with no answer to the question: How do reasons determine the normative result? Putting the question in terms of practical reasoning: How does one balance one's reasons in order to decide what to do? One answer to this question is suggested by Horty's account. We have already seen (Section 6) that Horty's account of our capacity to recognize context-dependent reasons saved the determination-by-reasons thesis from the threat of holism. But Horty's account is relevant to our concerns in this section as well. According to Horty's account (ideally), at the end of the process of recognizing reasons, all the reasons that favour one option are defeated so that all the undefeated reasons favour the second option and this option is the rational thing to do. In accounting for such ideal cases the notion of weight is dispensable, since there is no stage in this process that involves comparing weights. In less ideal cases, this process ends with one undefeated reason (or more) to do A and another undefeated reason (or more) to do B. In the literature on default logic, two strategies are suggested to treat the non-ideal cases, the first is called credulous (or bold) the second is called sceptic (or cautious). According to the first strategy, doing A and doing B are both rational, and reason requires us to choose arbitrarily; according to the second strategy, one should choose one option only when the other options are not supported at all. The point relevant for this paper is that whatever strategy we choose, in the less ideal cases doing A or doing B are described as rationally on a par. This means that one's situation in non-ideal cases is similar to Buridan's ass in that there is no place for rational deliberation anymore. If indeed the less ideal cases are like Buridan's cases then Horty provides a complete account of weighing reasons (practical deliberation) without weights of reasons. However, I suspect that the less ideal cases are not always similar to Buridan's case. For example, in the case of the job offer mentioned at the beginning of the paper, we can assume that your reason to take the third job (close to home) is not defeated. We can further assume that your reason against the job (the salary is not so high) is also not defeated. Horty's account implies that accepting the offer and rejecting it are rationally on a par; there is nothing else for reason to do. I believe that in contrast to the case of Buridan's ass your deliberation does not and should not end here. And this further deliberation is not accounted for by Horty's model. Accounting for this deliberation seems to call for the notion of weight. In deliberating whether to take the job or not you try to find out the weight of the two reasons involved and compare the weights. If my argument against DWT is sound, this cannot be the correct account, so we arrive again to the question raised at the beginning of this section: How does one balance one's reasons in order to decide what to do? I will address this question by looking at my own experience when deliberating whether to join the party with Alex and Carl. If I take the question seriously, I think about many things that I know about Alex and his behaviour in parties, I remember how I like his jokes and how in general I enjoy his company.25 Then, I take into account the fact that Carl will be there too and I imagine and think about an evening with both of them. I remember different facts about Carl and the interesting ideas that he often contributes to parties. I use all this knowledge about Alex and Carl and about other factors of the specific situation in trying to imagine the party and me in this party. In thinking about a party with Bruno and Carl, I enter a similar process and imagine the party with Bruno and Carl. In the end, I decide to go to the second party.26 A process, such as this, that involves experience and imagination is the core of practical deliberation, combining and comparing weights has nothing to do with it. Deciding what to do in a new situation is different from understanding a new sentence where the classic semantic theory provides a moderate account. 27 Deciding what to do in a new situation is more similar to cooking a new dish. In both cases, the outcome is determined by its components. The normative outcome is determined by the reasons involved and the gustatory outcome is determined by the tastes involved. In both cases the question: ‘how is the outcome determined by the components?’ (the demand for explanation of the determination), should be approached similarly. In both cases, it is wrong to approach the question by isolating a specific feature (weight or tastiness) of each component which accounts for the component's contribution to the outcome. In this respect, both cases are different from the semantic case where we can isolate a feature of each component (its reference) which accounts for the component's contribution to the outcome (truth value). Like every analogy, the culinary analogy is not complete. One important difference between the two cases is that weight of a reason is a theoretical notion while the tastiness of an ingredient is not. In particular, one can recognize the tastiness of an ingredient in isolation (prima facie tastiness) while, as was argued in Section 7, one cannot recognize the weight of a reason in isolation. Since the notion of the tastiness of an ingredient is intelligible without a theory, it is relatively simple to show that the thesis of determination by tastiness is false. The weight of a reason, on the other hand, is a theoretical notion, and the theory in which it is embedded is DWT. If DWT is false or empty the notion of weight loses its intelligibility. This makes the argument against DWT much more complex. So how can one know if the dish will be tasty before tasting it? And how can one know what to do before doing it? As mentioned above, there are many differences between these two questions, but it is the similarity that concerns me in this paper. In order to cook a tasty dish a good cook will not combine the ingredients according to their tastiness; instead he will use his experience and imagination to find a good combination of tastes (not of tastiness). In a similar way a good reasoner will not look at the weights of the reasons, instead he will use his experience and imagination to find a good combination of reasons (not of weights). Weights of reasons have no role in explaining practical reasoning. Footnotes 1 Schroeder (2007, 2011) suggests another theoretical role for the notion of weight. Reasons always come with weight and some feature is a reason only if its weight passes a certain threshold. For a powerful criticism of Schroeder's definition of the weight of a reason see Evers (2009). If Schroeder is right and reasons do have weights, the thesis of my paper should be reformulated as the claim that those weights cannot explain what we do when we weigh reasons. 2 As we will see later, Horty (2007) also objects to DWT, for him this thesis is too vague, and an account of reasoning should be more rigorous. I disagree, the problem with DWT is not that it is vague; the thesis is clear enough, the aim of this paper is to argue that it is false. 3 In saying that features of the situation have normative values I am not committed to objectivism; since a subjectivist might say that features of the situation have value only because we value it, or desire it, or believe it has value. 4 Value here is not to be confused with general values such as peace and justice; it is a theoretical entity that is ascribed to different outcomes according to the way they compare with other outcomes. 5 For example, Scanlon (1998:50–5) argues that since practical reasoning is not structured by weights (i.e., DWT is false), reasons cannot be based on desires and must have an objective base. 6 Even this claim is not completely true: ingredients such as salt and oil contribute to the taste of the whole more than their tastes. 7 This, of course, does not imply that the tastiness of the ingredients is irrelevant. 8 Kagan discusses this thesis in the context of moral reasoning and argues against it. 9 This is the way that we prove that the truth value of a counterfactual is not determined by the truth values of its components. We find two sentences A and B with the same truth value while the counterfactual composed of A and C has a different truth value than B and C. 10 I assume here that my preferences are rational, meaning that my preferences track the right balance of reasons. 11 My argumentative strategy here is similar to the particularist strategy in arguing for holism about reasons. But my dialectical aim is different; I am not arguing that weights vary with context. I am trying to establish that DWT and the claim that weights of reasons are invariant are in serious tension, such that one of them has to be rejected. If one rejects the first thesis (DWT) then the paper ends here and if one rejects the second (that weights of reasons are invariant), the argument of the paper starts here. 12 There are different theses in philosophy of language that are called holism. What I mean by holism here is context sensitivity: the semantic value of a word varies with the context of use. 13 This is what Dancy (2004: 190–8) does when he discusses the tension between holism about reasons and determination by reasons. 14 This distinction is different from the distinction between moderate and radical contextualism suggested by Cappelen & Lepore (2006: 429). Their distinction is based on the quantity of the terms that are context dependent, while my distinction is based on the quality of this dependence. 15 This claim should be qualified. Sometimes there is a limited range of interpretations of the parts. The listener then applies compositionality principles that determine a range of possible interpretations of the whole and chooses the one that is most likely to be true. I admit that this does happen sometimes but we do not want a semantic theory that works only sometimes. 16 Dancy himself does not take this route. 17 Default logic is generally studied in order to model theoretical reasoning (belief revision). I am not sure that the sceptic option is coherent in the practical case, since, in the practical case there is no analogue to being agnostic. This important asymmetry between believing and doing is not relevant for my argument. 18 I assume that the defeater was not itself defeated by a second-order defeater. This is where the account gets more technical. 19 I will discuss the less ideal cases in the last section. 20 The context might also often turn a reason for (or against) into a neutral reason a reason without direction. But notice that ‘reason’ with no direction is not a reason. 21 I do not assume that the notion of prima facie reason must necessarily have a role in the epistemology of recognizing reasons. What I do assume here is that, in order to understand the notion of prima facie reason, some abstraction similar to the isolation test is necessary. 22 It is true that sometimes we do say that R is a very strong (weighty) reason, but that is not an essential part of recognizing R as a reason per se. The fact that it happens sometimes helps to explain why we (mistakenly) think that reasons have weights. 23 One might try to save the notion of weights of reasons by following Star & Kearns (2013) in identifying practical reasons with evidence. Since strength of evidence is a well-accepted idea weight of a reason can be modelled on strength of evidence. I will briefly mention two objections to this direction. First, I believe that practical reasons should be sharply distinguished from theoretical reasons. Secondly, I believe that the arguments presented in this paper against weights of practical reasons can be directed against strength of evidence. But this is for another (more controversial) paper. 24 On the contrary, I even showed that this claim is not threatened by Holism about reasons. 25 I think about how I enjoy his company and not about how much I enjoy it. 26 Notice the similarity between this description and Dancy's description of moral thought: ‘Experience of similar cases can tell us what sort of thing to look out for, and the sort of relevance that a certain feature can have; in this way our judgement in a new case can be informed, though it is not forced or constrained, by our experience of similar cases in the past.’ (Dancy 2013). 27 I understand Travis (2008) as arguing that even in the semantic case the moderate account is inadequate and we should opt for the radical account there too. ‘When would it be so that someone saw an orange? Well, what are we to understand by seeing an orange (or seeing an orange)? The answer is to be gleaned neither from what seeing is, nor from what oranges are, as such.’ (Travis 2008:11) REFERENCES Broom J. ( 2004) ‘ Reasons’, in Wallace R. J. et al. (eds.) Reason and Value: Essays on the Moral Philosophy of Joseph Raz , 28– 55. Oxford: OUP. Google Scholar CrossRef Search ADS Cappelen H., Lepore E. ( 2006) ‘ Précis of Insensitive Semantics’, Philosophy and Phenomenological Research , LXIII: 424– 33. Dancy J. ( 2004) Ethics Without Principles . Oxford: Clarendon Press. Google Scholar CrossRef Search ADS Dancy J. ( 2013) ‘ Moral Particularism’, in: Edward, N.Z (ed.) The Stanford Encyclopedia of Philosophy , Fall edn. <https://plato.stanford.edu/archives/fall2013/entries/moral-particularism> accessed 24 April 2017. Evers D. ( 2009) ‘ Humean agent-neutral reasons?’, Philosophical Explorations: An International Journal for the Philosophy of Mind and Action , 12: 55– 67. Horty J.F. ( 2007) ‘ Defaults with priorities’, Journal of Philosophical Logic , 36: 367– 413. Google Scholar CrossRef Search ADS Kagan S. ( 1988) ‘ The additive fallacy’, Ethics , 99: 5– 31. Google Scholar CrossRef Search ADS Kearns S., Star D. ( 2013) ‘ Weighing Reasons’, Journal of Moral Philosophy , 10: 70– 86. Google Scholar CrossRef Search ADS Scanlon T.M. ( 1998) What We Owe to Each Other , Cambridge: Harvard University Press. Schroeder M. ( 2007) ‘ Weighting for a plausible Humean theory of reasons’, Noûs , 41: 138– 60. Google Scholar CrossRef Search ADS Schroeder M. ( 2011) ‘ Holism, weight, and undercutting’, Noûs , 45: 328– 44. Google Scholar CrossRef Search ADS Travis C. ( 2008) Occasion Sensitivity . Oxford: OUP. Google Scholar CrossRef Search ADS © The Author 2017. Published by Oxford University Press on behalf of The Scots Philosophical Association and the University of St Andrews. All rights reserved. For permissions, please e-mail: email@example.com
Philosophical Quarterly – Oxford University Press
Published: Jan 1, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera